The secondary-task method is sensitive when the two tasks compete for the same resource, and underestimates load when they don't

What it is

How sensitive the secondary-task technique is depends on whether the chosen secondary task competes with the primary task for the same kind of processing resource. When both need the same resource — both drawing on visual-spatial processing, for instance — secondary-task performance is highly sensitive to changes in primary-task load. When the secondary task draws on a resource the primary task barely touches — a visual tracking primary task paired with a purely auditory secondary task, say — the secondary task suffers little crowding out even when primary-task load is genuinely high, and the reading will systematically underestimate the true load.

Why it happens

This comes down to processing resources not being a single shared pool — processing of different kinds (visual, auditory, spatial, linguistic) is, to a considerable extent, independent. The secondary-task technique measures primary-task resource consumption precisely by having the two tasks "compete for space" within the same resource pool; only when they're genuinely competing for the same resource will the secondary task's performance suffer as the primary task consumes more of it. If the secondary task draws on an essentially non-overlapping resource, no matter how demanding the primary task gets, it can't claim the resource the secondary task depends on, so the secondary task keeps performing as usual. The resulting appearance of "stable load" is then just an artifact of picking the wrong type of secondary task, not evidence that the primary task is genuinely light.

Studying it

Choosing a secondary task requires first identifying which resource type the primary task mainly draws on — visual input, spatial processing, verbal/linguistic processing, manual response — and then selecting a secondary task that falls on that same resource type as much as possible, to keep the measurement sensitive. When the research goal is specifically to compare the load readings obtained from a "same-resource" secondary task versus a "different-resource" one on the same primary task, that comparison is itself a common confirmatory design used to work backward and identify which resource the primary task actually draws on. The methodological reminder: a "load isn't high" conclusion drawn from the secondary-task method is only trustworthy once resource overlap between the two tasks has been confirmed — otherwise the conclusion itself is unreliable.

Where it stops holding

When it's unclear which resource type the primary task draws on, or the primary task itself draws on a mix of several resource types, any single secondary task can only reflect the portion of resource consumption that overlaps with it. The resulting load reading is inherently a conservative (likely underestimated) estimate, and shouldn't be treated as a complete measurement of total load.

Related

• Same group: A9.08.1 Unchanged primary-task performance does not mean load hasn't increased — resources may simply have slack left · A9.08.2 The secondary-task method infers the primary task's remaining resources from the performance drop on an added task · A9.08.3 The secondary task's own presence changes how the primary task is performed — the measurement interferes with what it measures
• Adjacent: A9.10 Multiple Resource Theory
• Search terms: resource overlap · cross-modal secondary task · multiple resource theory